对于OpenCV-Python，OpenCV2.x和OpenCV3.x的函数使用方式有很大不同。网上很多教程都还是基于OpenCV2.x，此版本已经逐渐被弃用。
本教程针对特征点检测，分析OpenCV2.x和OpenCV3.x的不同之后，并重点介绍OpenCV3.x-Python的特征点检测。
Open2.x-Python 特征点检测方法
对于OpenCV2.x-Python，特征点检测及显示方法如下：# OpenCV2.x-Pythonfunction = cv2.Function_Name()keypoints = function.detect(img, None)img2 = cv2.drawKeyPoints(img, keypoints, color=(0,255,0))
其中Function_Name就是特征检测方法的函数名，如BRISK、FastFeatureDetector等。
比如，在OpenCV2.x-Python，想使用Fast来检测特征点，示例如下：
# OpenCV2.x-Pythonfast = cv2.FastFeatureDetector()keypoints = fast.detect(img, None)img2 = cv2.drawKeypoints(img, keypoints, color=(255,0,0))
Open3.x-Python 特征点检测方法
对于OpenCV3.x-Python，特征点检测及显示方法如下：
# OpenCV3.x-Python# 注意有_create()后缀function = cv2.Function_Name_create()keypoints = function.detect(img, None)# 注意显示之前要先将img2初始化img2 = img.copy()img2 = cv2.drawKeyPoints(img, keypoints, color=(0,255,0))
其中Function_Name就是特征检测方法的函数名，如BRISK、FastFeatureDetector等。
[注意1]：对于OpenCV3.x-Python，还要在Function_Name后加上_create后缀。其实这一点在opencv_doc中具体的函数python使用方法中已经注明了。
[注意2]：对于OpenCV3.x-Python，若要显示检测的特征点，需要初始化img2，才能正常显示。这里可以先使用img2 = img.copy()完成拷贝初始化。
下面就重点介绍OpenCV3.x-Python中的各种特征点检测方法的使用示例。
测试图像为标准的lena.png
AKAZE Feature Detection
#!/usr/bin/env python# -*- coding=utf-8 -*-# Summary: 使用OpenCV3.x-Python检测AKAZE特征点# Author: Amusi# Date: -03-17# Reference: /master/d8/d30/classcv_1_1AKAZE.htmlimport cv2import numpydef main():img = cv2.imread("lena.png")cv2.imshow('Input Image', img)cv2.waitKey(0)# 检测akaze = cv2.AKAZE_create()keypoints = akaze.detect(img, None)# 显示# 必须要先初始化img2img2 = img.copy()img2 = cv2.drawKeypoints(img, keypoints, img2, color=(0,255,0))cv2.imshow('Detected AKAZE keypoints', img2)cv2.waitKey(0)if __name__ == '__main__':main()
BRISK Feature Detection
#!/usr/bin/env python# -*- coding=utf-8 -*-# Summary: 使用OpenCV3.x-Python检测BRISK特征点# Author: Amusi# Date: -03-17# Reference: /master/de/dbf/classcv_1_1BRISK.htmlimport cv2import numpydef main():img = cv2.imread("lena.png")cv2.imshow('Input Image', img)cv2.waitKey(0)brisk = cv2.BRISK_create()keypoints = brisk.detect(img, None)# 必须要先初始化img2img2 = img.copy()img2 = cv2.drawKeypoints(img, keypoints, img2, color=(0,255,0))cv2.imshow('Detected BRISK keypoints', img2)cv2.waitKey(0)if __name__ == '__main__':main()
Fast Feature Detection
#!/usr/bin/env python# -*- coding=utf-8 -*-# Summary: 使用OpenCV3.x-Python检测FAST特征点# Author: Amusi# Date: -03-17# Reference: /master/df/d74/classcv_1_1FastFeatureDetector.htmlimport cv2import numpydef main():img = cv2.imread("lena.png")cv2.imshow('Input Image', img)cv2.waitKey(0)# -03-17 Amusi: OpenCV3.x FeatureDetector写法有变化# OpenCV2.x# fast = cv2.FastFeatureDetector()# keypoints = fast.detect(img, None)# OpenCV3.x# 注意有_create()后缀fast = cv2.FastFeatureDetector_create()keypoints = fast.detect(img, None)# 必须要先初始化img2img2 = img.copy()img2 = cv2.drawKeypoints(img, keypoints, img2, color=(0,255,0))cv2.imshow('Detected FAST keypoints', img2)cv2.waitKey(0)if __name__ == '__main__':main()
KAZE Feature Detection
#!/usr/bin/env python# -*- coding=utf-8 -*-# Summary: 使用OpenCV3.x-Python检测KAZE特征点# Author: Amusi# Date: -03-17# Reference: /master/d3/d61/classcv_1_1KAZE.htmlimport cv2import numpydef main():img = cv2.imread("lena.png")cv2.imshow('Input Image', img)cv2.waitKey(0)# 检测kaze = cv2.KAZE_create()keypoints = kaze.detect(img, None)# 显示# 必须要先初始化img2img2 = img.copy()img2 = cv2.drawKeypoints(img, keypoints, img2, color=(0,255,0))cv2.imshow('Detected KAZE keypoints', img2)cv2.waitKey(0)if __name__ == '__main__':main()
ORB Feature Detection
#!/usr/bin/env python# -*- coding=utf-8 -*-# Summary: 使用OpenCV3.x-Python检测ORB特征点# Author: Amusi# Date: -03-17# Reference: /master/db/d95/classcv_1_1ORB.htmlimport cv2import numpydef main():img = cv2.imread("lena.png")cv2.imshow('Input Image', img)cv2.waitKey(0)# 检测orb = cv2.ORB_create()keypoints = orb.detect(img, None)# 显示# 必须要先初始化img2img2 = img.copy()img2 = cv2.drawKeypoints(img, keypoints, img2, color=(0,255,0))cv2.imshow('Detected ORB keypoints', img2)cv2.waitKey(0)if __name__ == '__main__':main()
----------我是可爱的分割线----------
下面介绍属于nonfree的特征检测方法，如SIFT和SURF。
这些方法在opencv-contrib中，所以想要使用前，请卸载当前非contrib版本的opencv，即pip uninstall opencv-python后；再重新安装opencv-contrib-python，即pip install opencv-contrib-python
SIFT Feature Detection
#!/usr/bin/env python# -*- coding=utf-8 -*-# Summary: 使用OpenCV3.x-Python检测SIFT特征点# Author: Amusi# Date: -03-17# Reference: /master/d5/d3c/classcv_1_1xfeatures2d_1_1SIFT.htmlimport cv2import numpydef main():img = cv2.imread("lena.png")cv2.imshow('Input Image', img)cv2.waitKey(0)# 检测sift = cv2.xfeatures2d.SIFT_create()keypoints = sift.detect(img, None)# 显示# 必须要先初始化img2img2 = img.copy()img2 = cv2.drawKeypoints(img, keypoints, img2, color=(0,255,0))cv2.imshow('Detected SIFT keypoints', img2)cv2.waitKey(0)if __name__ == '__main__':main()
SURF Feature Detection
#!/usr/bin/env python# -*- coding=utf-8 -*-# Summary: 使用OpenCV3.x-Python检测SURF特征点# Author: Amusi# Date: -03-17# Reference: /master/d5/df7/classcv_1_1xfeatures2d_1_1SURF.htmlimport cv2import numpydef main():img = cv2.imread("lena.png")cv2.imshow('Input Image', img)cv2.waitKey(0)# 检测surf = cv2.xfeatures2d.SURF_create()keypoints = surf.detect(img, None)# 显示# 必须要先初始化img2img2 = img.copy()img2 = cv2.drawKeypoints(img, keypoints, img2, color=(0,255,0))cv2.imshow('Detected SURF keypoints', img2)cv2.waitKey(0)if __name__ == '__main__':main()
注：OpenCV3.x-Python与OpenCV2.x-Python有很多函数的用法不同，虽然网上教程大多参次不齐，但可以直接去官网查看最新的用法（官网即正义）
如果觉得《OpenCV-Python Feature2D 特征点检测(含SIFT/SURF/ORB/KAZE/FAST/BRISK/AKAZE)》对你有帮助，请点赞、收藏，并留下你的观点哦！
